Research On Coding Depth Algorithm Of Hevc Based On Machine Learning

Compared to picture and character, video can bring people a more real,more comprehensive experience, while video is the main promising content of the Internet in the future. With the popularity of mobile devices and the continuous improvement of video resolution, the existing transmission bandwidth is difficult to achieve the transmission of high-quality video. In order to solve the problem of transmission of high-resolution video, in January 2013, the International Video Joint Coding Group released a new generation of video compression coding standard,which called high efficiency video coding (HEVC) standard, it has become a research hotspot of video coding industry. Besides, domestic and international Experts and scholars focus on the fast coding depth algorithm.Machine learning is the hotspot of artificial intelligence research. In this paper, the depth selection algorithm is improved by combining machine learning techniques.The main work and innovations of this paper are as follows:Firstly, in this paper, the texture complexity of the image was quantified and the complexity relationship between the current coding unit and its four sub-blocks was deeply studied. In this paper, an intra coding unit depth decision algorithm based on K-means clustering algorithm was proposed, and the four-dimensional vector representing the sub-block complexity was taken as the input vector of the K-means algorithm.Meanwhile, the intra mode prediction process used a reduced number of prediction modes, and the shortcoming of the original intra-frame coding complexity had been improved. Experimental results indicate that the proposed algorithm can saves 50% of the intra-frame coding time, when compared to the standard HEVC test model 16.0.Secondly, this paper deeply analyzed the inter coding unit depth decision algorithm in HEVC, and fully studied the relationship between coding parameters and coding depth. In this paper, an inter coding unit depth selection algorithm based on support vector machine (SVM) is proposed, the optimal coding feature subset was selected by quantizing and cross-validating the coding features in different video sequences, which reduces the time consuming in the inter-frame coding process. Compared with the standard HEVC test model 16.0, experimental results reveal that the proposed algorithm reduces the inter-frame coding time by 46%.Thirdly, this paper conducted a survey of the current research progress of high-efficiency video coding at home and abroad, and studied these documents in detail. In this paper, based on the improved algorithm, the paper summarizes the current work, and prospects the future work.